WO1996001527A1 - Circuit de traitement de signaux - Google Patents

Circuit de traitement de signaux Download PDF

Info

Publication number
WO1996001527A1
WO1996001527A1 PCT/CA1995/000400 CA9500400W WO9601527A1 WO 1996001527 A1 WO1996001527 A1 WO 1996001527A1 CA 9500400 W CA9500400 W CA 9500400W WO 9601527 A1 WO9601527 A1 WO 9601527A1
Authority
WO
WIPO (PCT)
Prior art keywords
circuit
bit
serial
signal
floating point
Prior art date
Application number
PCT/CA1995/000400
Other languages
English (en)
Inventor
Gordon J. Reesor
Original Assignee
Mitel Corporation
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Mitel Corporation filed Critical Mitel Corporation
Priority to US08/765,205 priority Critical patent/US5991784A/en
Priority to JP50359096A priority patent/JP3375340B2/ja
Priority to EP95923162A priority patent/EP0769225B1/fr
Priority to DE69512413T priority patent/DE69512413T2/de
Publication of WO1996001527A1 publication Critical patent/WO1996001527A1/fr

Links

Classifications

    • H—ELECTRICITY
    • H03—ELECTRONIC CIRCUITRY
    • H03M—CODING; DECODING; CODE CONVERSION IN GENERAL
    • H03M7/00—Conversion of a code where information is represented by a given sequence or number of digits to a code where the same, similar or subset of information is represented by a different sequence or number of digits
    • H03M7/30—Compression; Expansion; Suppression of unnecessary data, e.g. redundancy reduction
    • H03M7/3002—Conversion to or from differential modulation
    • H03M7/3044—Conversion to or from differential modulation with several bits only, i.e. the difference between successive samples being coded by more than one bit, e.g. differential pulse code modulation [DPCM]
    • H03M7/3046—Conversion to or from differential modulation with several bits only, i.e. the difference between successive samples being coded by more than one bit, e.g. differential pulse code modulation [DPCM] adaptive, e.g. adaptive differential pulse code modulation [ADPCM]

Definitions

  • This invention relates to a circuit for applying a predetermined algorithm to an input signal.
  • Such a circuit may be, for example, a circuit for encoding or decoding speech samples using ADPCM (Adaptive Differential Pulse Code Modulation) .
  • ADPCM Adaptive Differential Pulse Code Modulation
  • ADPCM Adaptive Differential Pulse Code Modulation
  • CCITT recommendation G.721 (Melbourne 1988) describes the algorithm in detail as it pertains to the conversion of 64kb/s ⁇ -law or A-law PCM encoded speech to and from a 32kb/s compressed format.
  • ANSI recommendation Tl-303, and CCITT rec G.726 are similar documents with extensions to 40k/s and 24kb/s and 16kb/s bit rates.
  • the actual implementation of an ADPCM algorithm in a real time speech processing application may take various different forms ranging from a computer program, instruction code on a commercially available DSP chip, ASIC logic chip, or a custom integrated circuit.
  • This invention is concerned with implementation in the form of a custom chip.
  • implementation has been achieved in this form by employing microprocessor technology using a high speed clock to run a circuit consisting of a multiplexed ALU
  • Pair Gain is a method of increasing the number of subscribers that may simultaneously use a single analog twisted pair telephone line for independent two-way conversations.
  • A/D and D/A converters are used to digitize the speech channels into 64kb/s A-law or ⁇ - law PCM, after which one or more ADPCM devices are used to compress the 64kb/s streams to 32kb/s. These 32kb/s channels are then merged into a single ISDN Ubus transceiver device to transmit the digitized signals as one 144 kb/s (2B + D) base-band modem signal onto a single twisted pair copper cable. At the distant end of the cable, similar devices are used to reconvert the signals back to analog form to interface with separate analog telephones.
  • Pair Gain equipment is used in locations where the cost of installation of copper pair cable for additional telephone lines proves to be prohibitively expensive (or more expensive than the Pair Gain equipment) .
  • Pair Gain equipment circuitry be line powered (i.e. powered by DC on the line itself) , it is necessary that the power consumption of the devices used be minimized. There is a relationship between maximum line length possible with the Pair Gain equipment and the power consumption of the circuit. This means that the current consumption of the ADPCM device directly affects performance parameters of the Pair Gain product.
  • Cordless digital telephones that conform to the CT2 (Cordless Telephone 2) specification or other specifications use a codec to convert signals from analog to digital and back, as well as ADPCM compression to 32kb/s to reduce the bandwidth of the digital signal before transmission. Since the telephone set must be battery operated, the power consumption of the components will directly affect the number of hours of usage before the battery must be recharged. This means that the current consumption of the ADPCM device directly affects performance parameters of a cordless telephone product.
  • U.S. Patent No. 4,858,163 describes a serial arithmetic processor which is arranged to perform complex arithmetic functions of the ADPCM algorithm.
  • This patent (US 4,858,163) relates to a common means serial arithmetic processor (SAP) for efficiently performing certain selected complex arithmetic functions in the ADPCM algorithm, intended for use along with a micro- coded processor to implement the main body of ADPCM algorithm.
  • the micro-coded processor has a 16 bit bus connected to a RAM, an ALU core, and a 1024 X 29 bit ROM.
  • the Serial Arithmetic Processor is attached to this micro-coded processor via the multiplexed 16 bit bus.
  • the objects of this prior art patent are to provide an SAP comprising a common means to efficiently perform complex arithmetic functions such as LOG, ANTILOG, FLOATING POINT MULTIPLICATION and SIGNED MAG MULTIPLICATION.
  • the advantage of this design is stated to be a replacement for arithmetic functions which are burdensome to implement using the available instruction sets of presently available digital signal processing chips. This appears to be the only advantage, and the actual reasons for choosing serial logic circuitry are not well documented.
  • An object of the invention is to provide a circuit capable of implementing the ADPCM algorithm with reduced power consumption.
  • a circuit for applying a predetermined algorithm to an input signal signal processing means for processing said input signal in accordance with said predetermined algorithm, and means for outputting said processed signal, said signal processing means comprising distributed bit-serial logic circuits, for example in a parallel array, to implement said predetermined algorithm.
  • Bit serial logic circuitry has a substantially lower power consumption than parallel logic circuitry.
  • the present invention has a significant advantage over the prior art, particularly U.S. Patent No. 4,858,163, in that a micro-program processor architecture is not used.
  • all arithmetic functionality of the ADPCM algorithm is implemented as a parallel array of bit serial logic such that each arithmetic function is separate and not multiplexed (i.e. there are no common means for execution of different arithmetic functions) .
  • the ADPCM algorithm is actually hard wired into the connection of separate bit serial circuits.
  • Such an implementation has a more optimal way of performing arithmetic calculations from the point-of-view of low power consumption. The burden of fetching decoding and executing stored instructions is avoided. An example of why a stored micro-code implementation is less efficient will be described.
  • a functional step in a DSP algorithm requires a single bit to be set or cleared in a 16 bit binary register variable, which is stored in ram.
  • the micro-coded processor has to read this entire variable (all 16 bits) out of the RAM, move it to an ALU (arithmetic logic unit) , then set or clear the appropriate bit using logical instructions, and then move the result back to RAM.
  • This whole sequence of operations has the overhead of instruction fetch / decode / and execution from memory, as well as activity on the internal data bus for each of the 16 bits of the entire variable, as well as one or more ALU operations.
  • the corresponding logic designed specifically to implement this operation is as simple as directly setting or clearing a single flip-flop in a register. This therefore corresponds to a drastic reduction in total power drawn for this simple operation (when CMOS logic gates are used) .
  • the ⁇ QUANTIZER, INVERSE QUANTIZER, FORMAT CONVERSION, DIFF signal computation, TONE/TRANSITION detector, SCALE FACTOR, SPEED CONTROL filters are all implemented in serial hardware as opposed to software.
  • the FMULT operation which is the floating point multiply operation inside the ADAPTIVE PREDICTOR is probably the most arithmetically complex operation in the ADPCM algorithm.
  • the present invention does not use a parallel adder, and instead sums partial products together through an array of six serial adders to produce the final product in bit serial form which is shifted into shift register. In the CCITT FMULT definition, the product of the multiplication must be shifted according to the result of the addition of exponents.
  • Arithmetic operations such as additions, subtractions, 2's complement inversions, multiplications etc., require large amounts of logic gates if implemented in parallel. If instead, the operations are done serially, bit by bit beginning with the least significant bits, then the required amount of logic reduces significantly.
  • Preferably said predetermined algorithm is an ADPCM compression/decompression algorithm.
  • the implementation of the circuit is in the form of a Custom Integrated Circuit using standard cell 2 ⁇ m CMOS silicon technology.
  • the implementation follows the CCITT/ANSI recommendations for 32kb/s and 24kb/s ADPCM in terms of arithmetic processing; however, the use of distributed bit-serial logic implementation have resulted in substantially lower power consumption than the prior art, namely in the order of 10-12 mwatts for one encoder and one decoder using a 5v power supply, or 3-5 mWatts using a 3V power supply.
  • Figure 1 is a block diagram of an ADPCM encoder operating according to CCITT standard G.726
  • Figure 2 is a block diagram of an ADPCM decoder operating according to CCITT standard G.726;
  • Figure 3 is a circuit of a bit serial adder in accordance with one embodiment of the invention
  • Figure 4 is a circuit of a bit serial subtracter in accordance with one embodiment of the invention
  • Figure 5 is a block diagram of an adaptive predictor as defined by CCITT specifications
  • Figure 6 is an ADDC implementation circuit
  • Figure 7 is a floating point multiplier unit (FMULT) ;
  • Figure 8 is a block diagram of an FMULT floating point converter
  • Figure 9 shows a Prediction Coefficient Update Circuit (UPB, XOR, TRIGB) ;
  • Figure 10 is a floating point conversion circuit (FLOATA) ;
  • Figure 11 is a timing chart for the circuit of Figure 10; and Figure 12 shows a second floating point converter circuit (FLOATB) .
  • FLOATB floating point converter circuit
  • Figure 1 shows a block diagram of a circuit implementing the ADPCM algorithm as defined in CCITT specifications.
  • the 64kb/s ADPCM input stream s (k) is input to a ADPCM format converter 1 connected to difference signal unit 2.
  • the difference signal d(k) is fed to adaptive quantizer 3, which produces an output signal I (k) that is input to inverse adaptive quantizer 4 producing an output d g (k) input to adaptive predictor 5.
  • the outputs of inverse adaptive quantizer 4 and adaptive predictor 5 are respectively applied as inputs to reconstructed signal calculator 6 whose output s r (k) is applied to adaptive predictor 5.
  • inverse adaptive quantizer d (] ) is also applied to tone transition detector 7, along with an output of adaptive predictor 5, and the output of tone and transition detector 7 is applied to adaptive speed control unit 8, in turn connected to quantizer scale factor and adaptation unit 9.
  • the ADPCM output appears at the input to inverse adaptive quantizer 4.
  • the encoder circuit Figure 1 is specified in document CCITT G.726, to which the reader is referred. Throughout this specification the processing variables will be identified using the same terminology as is employed in this document .
  • Figure 2 shows a decoder circuit which receives at its input an ADPCM signal and provides at its output a PCM signal s ⁇ ⁇ [(k) .
  • the individual component of the circuits of Figure 2 are generally similar to those of Figure 1, and like reference numbers are employed where appropriate.
  • the output of the reconstructed signal calculator 6 is applied to an output PCM format conversion circuit 10, which in turn is connected to a synchronous coding adjustment circuit 11. This is circuit is also described in detail in CCITT document G.726.
  • the bit serial adder comprises an exclusive OR gate 20 having inputs A and B and an output connected to one input of an exclusive OR gate 21.
  • the inputs of AND gate 22 are also connected to respective inputs A and B, and the inputs of AND gate 23 are respectively connected to output of exclusive OR gate 20 and a second input of exclusive OR gate 21, which in turn is connected to the Q output of bistable flip flop 24.
  • the outputs of AND gates 22 are connected through NOR gate 25 to the D input of flip flop 24.
  • Arithmetic operations such as additions, subtractions, 2's complement inversions, multiplications etc., require large amounts of logic gates if implemented in parallel. If instead, the operations are done serially, bit by bit beginning with the least significant bits, then the required amount of logic reduces significantly.
  • the XOR gates 20, 21 perform the single bit 2's complement addition, and the carry bit generated by the AND /OR combination is latched and used during the next single bit addition. Initialization of the latched carry bit C is done using the signal PRESET before the first addition of the LSB'S.
  • PRESET on the flip-flop has changed from a SET to a RESET function, which effectively causes the first carry bit C to be a ONE, and the B input bits are all complemented resulting in a 2's complement addition of (A plus the one's complement of B plus ONE) .
  • This methodology can be easily extended to implement any arithmetic or logical function, including multiplication. Delays or register storage are implemented using shift registers, which are clocked only when needed.
  • the adaptive predictor comprises a parallel array of six FMULT units 30 connected to respective adders 31, each FMULT unit 30 producing serial bit streams B1 to WB6.
  • Floating point converter 36 shown in more detail in Figure 10, provides an input to the array.
  • a second parallel array of two floating point multipliers 30 receives an input from floating point converter 37, shown in more detail in Figure 12, and outputs signals A1, WA2. These bit streams are summed in the adders 31, which produce outputs SE and SEZ as required by the ACCUM operation defined in the CCITT specifications.
  • the first six FMULT units 30, producing outputs B1 to B6, are connected to combined XOR, UPB and TRIGB predictor units 32 (shown in more detail in Figure 9) , whereas the remaining FMULT units 30 producing outputs A1, A2 are connected to predictor units 33.
  • the circuit of Figure 5 generally operates in accordance with the CCITT G.726 ADPCM.
  • FIG 6 shows a circuit in accordance with the invention, which implements ADDC operation (block 34 in Figure 5) as well as PK delays (inputs to predictors 33 in Figure 5) as defined in CCITT G.726.
  • serial streams SEZ and DQ are summed in serial adder 40 to form a resulting signal DQSEZ which is latched into flip flops 41 at the appropriate time using clock signal ENl, which is generated elsewhere in the chip.
  • the outputs of the flip flops 41 form signals PK0, PK1, and PK2 respectively.
  • RS latch 42 generates an output SIGPK by first clearing SR latch output low using reset signal START which occurs before serial adder 40 begins its calculation.
  • Each bit of the resulting signal DQSEZ is gated with clock CK1 using AND gate 43 to deglitch the bit signals to form a set signal to SR latch 42.
  • FIG. 7 shows in detail a floating point multiplier block 30 (FMULT) in accordance with the invention.
  • DQn signals shift serially through Shift Register 50 to implement to the DQO to DQ6 delay line.
  • Shift Register 50 is tapped off in parallel to provide the components of the DQn (Quantized Difference) signal to the floating point multiplier.
  • These components are the six bit wide DQnMANT, the four bit wide DQnEXP, and the single bit DQnS, and are available to be used by the floating point multiplier when Shift Register 50 is static, i.e. when SCLK is not active.
  • the other input to the floating point multiplier is from 13 bit linear magnitude BnMAG and sign bit BnS.
  • BnMAG is shifted serially into floating point converter 51 and converted into six bit wide BMANT signal and four bit wide BnEXP signal .
  • BnMANT and DQnMANT are gated together through AND gate array 52 and serial adder array 53 to form the result BnMANT, which is a 12 bit serial bit stream representing the mantissa of the product.
  • the six bit bus DQnMANT is a collection of static signals representing a multiplicand, but the six bit bus BnMANT is not a static bus and is actually a collection of shifted serial bit streams generated by the floating point converter 51.
  • the following shows the serial bit streams as defined on each signal of BnMANT, shifted LSB first.
  • shift register 54 is an 8 bit register.
  • the value held in shift register 54 is held for a period of time and then shifted out at an appropriate time to generate 16 bit serial signal WBn.
  • AND gate 56 is used to remove unwanted bits from WBnMAG which is then converted from a magnitude format to a 2's complement format by MUX circuit 57 to generate WBn in serial fashion.
  • MUX 57 is used to select either the magnitude WBnMAG or the 2 ' s complement of WBnMAG generated by serial complement circuit 58.
  • MUX 57 is controlled by XOR gate 59, the inputs of which are the sign bits of the multiplier BnS and multiplicand DQnS.
  • exponent adder 60 adds BnEXP to DQnEXP and produces a 5 bit result WBnEXP.
  • WBnEXP is used as the preset value loaded into a 5 bit down counter 61.
  • the three most significant bits of this down counter are decoded by logic circuit 55 to generate signal CKEN. If a value greater than or equal to 11 is loaded into the counter then the clock to the down counter becomes enabled by CKEN. When the counter reaches a count of 11 then CKEN will stop the clock to the counter freezing the count at 11.
  • Signal EN2SCOMP is also generated and conditioned by circuit 55 to produce the clock WCLK used to clock shift register 54.
  • the delay before WCLK is started is dependent upon the value of WBnEXP produced by the summation of exponents 60, and the number of clock cycles before the down counter reaches a count of 11. This implements the required shifting to scale WBnMAG as per CCITT G.726.
  • the Floating Point Converter block 51 is shown in Figure 8.
  • a 6 bit shift register 70 is first cleared by the START signal, which also initializes through OR gate 71, the loadable shift register 72, to the binary value "100000", shift register 73 of all zeros, as well as counter 74 to a count of 13.
  • Serial input signal BnMAG is shifted into least significant bit first using a clock signal COUNTCLK.
  • the Predictor Coefficient Update (UPB) block is shown in Figure 9.
  • This circuit performs an adaptive coefficient update for the predictor filter.
  • the outputs of this circuit provide the multiplier input to an FMULT floating point multiplier circuit shown in Figure 7.
  • the 16 bit 2 ' s complement representation of the filter coefficient Bn is stored in shift register 80.
  • clock signal SHIFTCLK becomes active shifting the register contents through serial adders 81 and 82.
  • Signal Bn>>8 is a signal tapped from the 8th flip-flop of the shift register and represents the value of Bn shifted right by 8 bits.
  • Latch 83 allows signal Bn>>8 to pass through transparently until the most significant bit (the sign bit) of Bn>>8 is present at the output of latch 83, the latch enable EN3 (from elsewhere in the chip) is then changed to logic "0" causing the sign bit of Bn>>8 to become latched so as to extend the sign bit through subsequent bit periods so as to effect a 2's complement sign bit extension.
  • AND gate 84 clears serial bit stream BnP to all zeroes when input signal TR is asserted to logic "1' to implement the TRIGB definition as per CCITT G.726.
  • XOR gate 85 implements the XOR operation on input sign bits DQS and DQnS to generate signal "Un" as per
  • Logic gates 86 along with input strobe pulses STRB1, STRB2, and STRB3, generate signal UGBn which is one of 3 possible 2 ' s complement serial codes (+80hex, -80hex, or 0) .
  • signals BnMAG and BnS are produced and transmitted to the FMULT circuit.
  • the sign bit of Bn (labeled as Bn>>15) is tapped-off the shift register at the rightmost bit position. This signal is latched at the appropriate time by signal EN3 to latch 86 to hold the sign bit BnS.
  • Bn>>2 is also tapped-off the shift register two positions from the left most end of the register and passed through serial circuit 87 which converts the 2's complement representation of Bn>>2 to signed magnitude representation BnMAG.
  • serial circuit 87 which converts the 2's complement representation of Bn>>2 to signed magnitude representation BnMAG.
  • the operation of this conversion circuit is similar to the one in the FMULT circuit.
  • the FLOATA block shown in Figure 10 converts a 15 bit signed magnitude DQMAG signal along with sign bit DQS to floating point representation for use by the FMULT floating point multipliers.
  • the inputs and outputs of this circuit are serial format least significant bit first.
  • Shift register 90 is first cleared by the START signal, which also initializes through OR gate 91, the loadable shift register 92, and 4 bit down counter 93.
  • Serial input signal DQMAG is shifted into shift register 90 least significant bit first using clock signal CLK1.
  • CLK1 clock signal
  • the loadable shift register 92 will now contain the previous 5 bits in the DQMAG stream before the logic "1" was detected, along with a "1" in the most significant bit position. Since a multiple number of occurrences of "1” may occur in the DQMAG input stream, this load process is repeated until the last occurrence of a "1” is found. Every subsequent "0" in the DQMAG stream will have no effect upon the circuit except that the down counter 93 will decrement by 1 count. After the last bit of DQMAG has been clocked into shift register 90, CLK2 becomes active (CLK1 remains active) and 4 bit down counter 93 stops counting and switches its function to that of a shift register; the change in function being indicated by control input COUNT/SHIFT generated elsewhere on the chip.
  • the value left in the down counter is the desired exponent value DQEXP, which along with input sign bit DQS becomes serially connected to shift register 92 and the final result DQO is shifted out serially using CLK1 and CLK2.
  • the input and output signals are shown in Figure 11.
  • the FLOATB block shown in Figure 12 converts a 16 bit 2's complement SR signal to floating point representation for use by the FMULT floating point multipliers.
  • the inputs and outputs of this circuit are serial format least significant bit first.
  • the operation of this circuit is identical to that of the FLOATA circuit previously described, with the exception of the inclusion of a 2's complement to signed magnitude conversion circuit 100, 101, 102.
  • circuits described show how distributed serial logic can be applied to implement the ADPCM algorithm as defined in the CCITT specifications with economy of power and logic circuitry. It will be apparent to a person skilled in the art that the invention will find many other applications in the field of digital signal processing.
  • the circuits are particularly advantageous because of their very low power consumption, which is extremely valuable in small portable battery powered devices, such as telephone handsets.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)
  • Analogue/Digital Conversion (AREA)
  • Complex Calculations (AREA)
  • Communication Control (AREA)
  • Logic Circuits (AREA)

Abstract

Un circuit pour appliquer un algorithme prédéterminé à un signal d'entrée comprend une entrée pour recevoir le signal d'entrée, un dispositif de traitement des signaux pour traiter le signal d'entrée conformément à l'algorithme prédéterminé et un dispositif pour sortir le signal traité, le dispositif pour traiter le signal comprenant des circuits logiques à constantes réparties en séries par bits pour mettre en oeuvre l'algorithme prédéterminé.
PCT/CA1995/000400 1994-07-06 1995-07-06 Circuit de traitement de signaux WO1996001527A1 (fr)

Priority Applications (4)

Application Number Priority Date Filing Date Title
US08/765,205 US5991784A (en) 1994-07-06 1995-07-06 Signal processing circuit
JP50359096A JP3375340B2 (ja) 1994-07-06 1995-07-06 信号処理回路
EP95923162A EP0769225B1 (fr) 1994-07-06 1995-07-06 Circuit de traitement de signaux
DE69512413T DE69512413T2 (de) 1994-07-06 1995-07-06 Signalverarbeitungsschaltung

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CA2,127,520 1994-07-06
CA002127520A CA2127520C (fr) 1994-07-06 1994-07-06 Circuit de traitement de signaux

Publications (1)

Publication Number Publication Date
WO1996001527A1 true WO1996001527A1 (fr) 1996-01-18

Family

ID=4153965

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CA1995/000400 WO1996001527A1 (fr) 1994-07-06 1995-07-06 Circuit de traitement de signaux

Country Status (6)

Country Link
US (1) US5991784A (fr)
EP (1) EP0769225B1 (fr)
JP (1) JP3375340B2 (fr)
CA (1) CA2127520C (fr)
DE (1) DE69512413T2 (fr)
WO (1) WO1996001527A1 (fr)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0801377A2 (fr) * 1996-04-12 1997-10-15 Nec Corporation Procédé et appareil pour coder un signal

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8301803B2 (en) * 2009-10-23 2012-10-30 Samplify Systems, Inc. Block floating point compression of signal data
CN112435663A (zh) * 2020-11-11 2021-03-02 青岛歌尔智能传感器有限公司 命令语音管理方法、装置、设备及介质

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4858163A (en) * 1987-07-02 1989-08-15 General Datacomm, Inc. Serial arithmetic processor
EP0334714A1 (fr) * 1988-03-14 1989-09-27 France Telecom Codeur différentiel à filtre prédicteur auto adaptatif et décodeur utilisable en liaison avec un tel codeur
GB2218548A (en) * 1988-04-12 1989-11-15 Texas Instruments Ltd Digital signal processor

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO1986002474A1 (fr) * 1984-10-16 1986-04-24 The Commonwealth Of Australia Care Of The Secretar Multiplicateur cellulaire pipeline en serie a virgule flottante

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4858163A (en) * 1987-07-02 1989-08-15 General Datacomm, Inc. Serial arithmetic processor
EP0334714A1 (fr) * 1988-03-14 1989-09-27 France Telecom Codeur différentiel à filtre prédicteur auto adaptatif et décodeur utilisable en liaison avec un tel codeur
GB2218548A (en) * 1988-04-12 1989-11-15 Texas Instruments Ltd Digital signal processor

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
SAMI ALY: "24-Channel 32 Kb/s ADPCM Transcoder Using the CCITT Recommendation g.721", INTERNATIONAL CONFERENCE ONE ACOUSTICS, SPEECH AND SIGNAL PROCESSING ICASSP, vol. 1, TOKYO, pages 349 - 352 *
SHANBHAG N R ET AL: "A HIGH-SPEED ARCHITECTURE FOR ADPCM CODEC", PROCEEDINGS OF THE INTERNATIONAL SYMPOSIUM ON CIRCUITS AND SYSTEMS, SAN DIEGO, MAY 10 - 13, 1992, vol. 3 OF 6, 10 May 1992 (1992-05-10), INSTITUTE OF ELECTRICAL AND ELECTRONICS ENGINEERS, pages 1499 - 1502, XP000338228 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0801377A2 (fr) * 1996-04-12 1997-10-15 Nec Corporation Procédé et appareil pour coder un signal
EP0801377A3 (fr) * 1996-04-12 1998-09-23 Nec Corporation Procédé et appareil pour coder un signal
US5857168A (en) * 1996-04-12 1999-01-05 Nec Corporation Method and apparatus for coding signal while adaptively allocating number of pulses

Also Published As

Publication number Publication date
JP3375340B2 (ja) 2003-02-10
US5991784A (en) 1999-11-23
EP0769225B1 (fr) 1999-09-22
EP0769225A1 (fr) 1997-04-23
CA2127520A1 (fr) 1996-01-07
JPH10504428A (ja) 1998-04-28
DE69512413T2 (de) 2000-02-03
CA2127520C (fr) 2001-01-16
DE69512413D1 (de) 1999-10-28

Similar Documents

Publication Publication Date Title
US6243729B1 (en) Digital finite-impulse-response (FIR) filter with a modified architecture based on high order Radix-N numbering
US5941940A (en) Digital signal processor architecture optimized for performing fast Fourier Transforms
US4598266A (en) Modulo adder
Shen et al. Low-power multipliers by minimizing switching activities of partial products
US3988606A (en) Digital filter device for processing binary-coded signal samples
JP2998551B2 (ja) 回路領域の少ないディジタル積分器およびそれを用いたアナログ・ディジタル変換器
US6408102B1 (en) Encoding/decoding device
EP0769225B1 (fr) Circuit de traitement de signaux
McCanny et al. Hierarchical VHDL libraries for DSP ASIC design
US5563816A (en) High-resolution digital filter
JPS6136415B2 (fr)
JPH05218801A (ja) 間引きフィルタ
JP3418595B2 (ja) 三値ディジタル回路
US5420584A (en) Data converter with barrel shifter
JPH0126207B2 (fr)
JP3576208B2 (ja) Ｄｓｐによるビットレート切換機能付データ圧縮方法および装置
US5765013A (en) Digital signal processor
CN112988111B (zh) 一种单比特乘法器
KR0155797B1 (ko) 디지탈 신호 처리장치
Hatem et al. VLSI architecture of QMF for DWT integrated system
Roy et al. An efficient two's complement systolic multiplier for real-time digital signal processing
McCanny et al. Rapid design of DSP ASIC cores using hierarchical VHDL libraries
Mok et al. A VLSI decimation filter for sigma-delta A/D converters
CA1127307A (fr) CONVERTISSEUR DE LOI .mu. EN VIRGULE FLOTTANTE
Vergos A 200-mhz rns core

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): DE GB JP US

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): AT BE CH DE DK ES FR GB GR IE IT LU MC NL PT SE

DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
121 Ep: the epo has been informed by wipo that ep was designated in this application
WWE Wipo information: entry into national phase

Ref document number: 1995923162

Country of ref document: EP

WWP Wipo information: published in national office

Ref document number: 1995923162

Country of ref document: EP

REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

WWE Wipo information: entry into national phase

Ref document number: 08765205

Country of ref document: US

WWG Wipo information: grant in national office

Ref document number: 1995923162

Country of ref document: EP